Complex Environment To Understand The Terrain Classification In Research And Analysis

Terrain classification under complex environment has its extremely important significance in autonomous navigation. At present,some approaches of the terrain classification could not completely meet the requirement of outdoor autonomous navigation. In this paper,by taking the distribution of 3-D point cloud as the research object,study on the classification techniques based on the spatial characteristics of ladar data,carry out the experiments and analysis the experiment result.The object of the algorithm is scattered point clouds,or called unorganized cloud points. That is,the data of the point clouds include only the space coordinate of the point,no other information. In the process of the extraction of the special distribution of point clouds,we need the K-nearest neighbours. This paper uses the parameter adjustable single axes searching arithmetic algorithm. Then use the coordinate covariance matrix and direction covariance matrix to extract the feature vector.Point clouds of different terrains have different spatial distribution. Extract the spatial geometric feature to train the classifier. Then use the classifier for terrain classification. This paper uses the GMM&Bayesian classifier and the support vector machines classifier. Then statistic and analysis the experimental data and experimental results.In applications,this paper summarizes the current development of terrain classification algorithms. Use Riegl LMS-Z210i ground-based laser scanner to obtain point clouds. Then classify the point clouds into different terrains. Through comparative analysis,summarizes the advantages and limitations of ladar-based terrain classification approaches.